#functions that we need
#install.packages('bnlearn')
library(bnlearn)
# define a function that takes a dataframe of returns, a type of test and a fixed variable
# it returns a dataframe with p-values and correlations for every pair of stocks given the fixed variable
ci_test_pairs <- function(df, test, fixed) {
# get the names of the columns that are not the fixed variable
stocks <- names(df)[names(df) != fixed]
# create an empty dataframe to store the results
results <- data.frame()
# loop over all pairs of stocks
for (i in 1:(length(stocks) - 1)) {
for (j in (i + 1):length(stocks)) {
# get the names of the pair
x <- stocks[i]
y <- stocks[j]
# perform the conditional independence test given the fixed variable
test_result <- ci.test(x, y, fixed, data = df, test = test)
# extract the p-value and the correlation from the test result
p_value <- test_result$p.value
correlation <- cor(df[[x]], df[[y]])
# append a row to the results dataframe with the pair names, p-value and correlation
results <- rbind(results, data.frame(x = x, y = y, p_value = p_value, correlation = correlation))
}
}
# return the results dataframe
return(results)
}
correlation_threshold <- function(cor_matrix, threshold) {
# Get the lower triangle of the correlation matrix
cor_lower <- cor_matrix[lower.tri(cor_matrix)]
# Get the indices of the lower triangle of the correlation matrix
cor_indices <- which(lower.tri(cor_matrix), arr.ind = TRUE)
# Create a data frame with the pairs of variables and their correlations
df <- data.frame(x = rownames(cor_matrix)[cor_indices[, 1]],
y = colnames(cor_matrix)[cor_indices[, 2]],
edge = ifelse(abs(cor_lower) >= threshold, 1, 0))
return(df)
}
p_value_eval<- function(x,thresh){
if(x<thresh){
ret=1
}
else{
ret=0
}
return(ret)
}
grapher<-function(example1,compl){
library(dplyr)
library(igraph)
#filtered only the edges that contain 1
edges1 <- filter(example1, edge == "1")
edges1
#plotted the graph
g <- graph_from_data_frame(edges1, directed = FALSE)
if(compl==TRUE){
g<- complementer(g)
}
plot(g, vertex.label = V(g)$names, edge.label = E(g)$weight)
}
#reading the data and getting rid of HSI
df<- read.csv("/users/amirreza/Desktop/returns_data.csv")
df<- df[,-12]
#different slices corresponding to federal reserve interest rates hiking/cutting/pausing
df1<- df[ df$Date> "2001-11-01" & df$Date < "2001-12-31",-1 ]
df2<- df[ df$Date> "2002-01-01" & df$Date < "2002-10-31",-1 ]
df3<- df[ df$Date> "2002-11-01" & df$Date < "2004-06-31",-1 ]
df4<- df[ df$Date> "2004-07-01" & df$Date < "2006-08-31",-1 ]
df5<- df[ df$Date> "2006-08-31" & df$Date < "2007-08-01",-1 ]
df6<- df[ df$Date> "2007-08-01" & df$Date < "2008-05-01",-1 ]
df7<- df[ df$Date> "2008-05-01" & df$Date < "2008-09-31",-1 ]
df8<- df[ df$Date> "2008-09-31" & df$Date < "2009-01-01",-1 ]
df9<- df[ df$Date> "2009-01-01" & df$Date < "2015-12-31",-1 ]
df10<- df[ df$Date> "2016-01-01" & df$Date < "2019-07-31",-1 ]
df11<- df[ df$Date> "2019-07-31" & df$Date < "2020-05-01",-1 ]
df12<- df[ df$Date> "2020-05-01" & df$Date < "2022-02-01",-1 ]
dff = list(df1,df2,df3,df4,df5,df6,df7,df8,df9,df10,df11,df12)
results=list()
for (i in 1:12){
res<-ci_test_pairs(df = dff[[i]],fixed="DXY")
results[[i]]<- res
}
#the result of the test is available as a list, each element corresponds to a time-slice
df<- read.csv("/users/amirreza/Desktop/returns_data_classified.csv")
df<- df[,-12]
#different slices corresponding to federal reserve interest rates hiking/cutting/pausing
df1<- df[ df$Date> "2000-11-01" & df$Date < "2001-12-31",-1 ]
df2<- df[ df$Date> "2002-01-01" & df$Date < "2002-10-31",-1 ]
df3<- df[ df$Date> "2002-11-01" & df$Date < "2004-06-31",-1 ]
df4<- df[ df$Date> "2004-07-01" & df$Date < "2006-08-31",-1 ]
df5<- df[ df$Date> "2006-08-31" & df$Date < "2007-08-01",-1 ]
df6<- df[ df$Date> "2007-08-01" & df$Date < "2008-05-01",-1 ]
df7<- df[ df$Date> "2008-05-01" & df$Date < "2008-09-31",-1 ]
df8<- df[ df$Date> "2008-09-31" & df$Date < "2009-01-01",-1 ]
df9<- df[ df$Date> "2009-01-01" & df$Date < "2015-12-31",-1 ]
df10<- df[ df$Date> "2016-01-01" & df$Date < "2019-07-31",-1 ]
df11<- df[ df$Date> "2019-07-31" & df$Date < "2020-05-01",-1 ]
df12<- df[ df$Date> "2020-05-01" & df$Date < "2022-02-01",-1 ]
dff_count = list(df1,df2,df3,df4,df5,df6,df7,df8,df9,df10,df11,df12)
results_count=list()
for (i in 1:12){
res_count<-ci_test_pairs(df = dff_count[[i]],fixed="DXY")
results_count[[i]]<- res_count
}
#plotting the graphs for inverse of correlation matrix
#inver<- list()
#for(i in 1:12){
# inver[[i]]<- solve(cor(dff[[i]]))
#}
set.seed(123)
for(i in 1:12){
grapher(correlation_threshold(cor(dff[[i]]),threshold = 0.08),compl = TRUE)
title(paste("Marginal correlation graphs for the ", i, "th slice"))
}
##
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
##
## filter, lag
## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union
##
## Attaching package: 'igraph'
## The following objects are masked from 'package:dplyr':
##
## as_data_frame, groups, union
## The following objects are masked from 'package:bnlearn':
##
## as.igraph, compare, degree, subgraph
## The following objects are masked from 'package:stats':
##
## decompose, spectrum
## The following object is masked from 'package:base':
##
## union
#plotting the graph for conditional independence test on the continuous returns
graphss <- list()
for(i in 1:12){
l<- lapply(results[[i]]$p_value, p_value_eval, thresh=0.0001)
results[[i]]$edge<- as.integer(l)
graphss[[i]] <- results[[i]][c("x","y","edge")]
grapher(graphss[[i]],compl = TRUE)
title(paste("CI test of continous returns for the " ,i, "th slice, thresh=0.0001"))
}
#plotting the graph for the conditional independence test on the count(1,0,-1) data
for(i in 1:12){
set.seed(123)
l<- lapply(results_count[[i]]$p_value, p_value_eval, thresh=0.0001)
results_count[[i]]$edge<- as.integer(l)
graphss[[i]] <- results_count[[i]][c("x","y","edge")]
grapher(graphss[[i]],compl = TRUE)
title(paste("CI test of counts of returns for the " ,i, "th slice, thresh=0.0001"))
}
graphss <- list()
for(i in 1:12){
l<- lapply(results[[i]]$p_value, p_value_eval, thresh=0.0001)
results[[i]]$edge<- as.integer(l)
graphss[[i]] <- results[[i]][c("x","y","edge")]
grapher(graphss[[i]],compl = TRUE)
title(paste("CI test of continous returns for the " ,i, "th slice, thresh=0.0001"))
}
#plotting the graph for the conditional independence test on the count(1,0,-1) data
for(i in 1:12){
set.seed(123)
l<- lapply(results_count[[i]]$p_value, p_value_eval, thresh=0.0001)
results_count[[i]]$edge<- as.integer(l)
graphss[[i]] <- results_count[[i]][c("x","y","edge")]
grapher(graphss[[i]],compl = TRUE)
title(paste("CI test of counts of returns for the " ,i, "th slice, thresh=0.0001"))
}